[BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++

linux-gcc.vger.kernel.org archive mirror
 help / color / mirror / Atom feed

* [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
@ 2004-12-22  3:42 Denis Zaitsev
  2004-12-22  4:06 ` Andrew Pinski
  0 siblings, 1 reply; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-22  3:42 UTC (permalink / raw)
  To: gcc, linux-gcc

I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
I mean either of these flags is present in both the exported shell
variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
from gcc/Makefile.  Both the builds was successfull.  And these are
some results of 'make check':


-fno-omit-frame-pointer:

                === libstdc++ Summary ===

# of expected passes            2720
# of unexpected failures        1
# of unexpected successes       1
# of expected failures          6


-fomit-frame-pointer:

                === libstdc++ Summary ===

# of expected passes            2364
# of unexpected failures        357
# of expected failures          7


So, the difference seems to be just huge.  Does it mean some serious
problems GCC has in the hard-fp-less area for x86?  (BTW, it
traditionally has some...)  The tests results for gcc and g++ are
different too, but that differencies are just miserable compared with
those listed.

Other details about the build and check environment:

CFLAGS, CXXFLAGS, BOOT_CFLAGS:

        -pipe
        -O3
        -fno-strength-reduce
        -fno-inline-functions
        -fno-align-functions
        -fno-align-loops
        -fno-align-jumps
        -fno-ident
        -f[no]omit-frame-pointer
        -D__USE_STRING_INLINES
        -mpreferred-stack-boundary=2

configure's args:

        --enable-shared
        --enable-static
        --disable-debug
        --disable-profile
        --disable-nls
        --disable-libtool-lock
        --disable-rpath
        --disable-largefile
        --with-gnu-as
        --with-gnu-ld
        --enable-threads=posix
        --enable-__cxa_atexit
        --enable-languages=c,c++,objc
        --enable-version-specific-runtime-libs
        --enable-clocale
        --enable-objc-gc
        --disable-checking
        --with-gc=simple
        --with-system-zlib
        --with-tune=pentium

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  3:42 [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++ Denis Zaitsev
@ 2004-12-22  4:06 ` Andrew Pinski
  2004-12-22  4:34   ` Denis Zaitsev
                     ` (3 more replies)
  0 siblings, 4 replies; 8+ messages in thread
From: Andrew Pinski @ 2004-12-22  4:06 UTC (permalink / raw)
  To: Denis Zaitsev; +Cc: gcc, linux-gcc


On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:

> I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
> I mean either of these flags is present in both the exported shell
> variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
> from gcc/Makefile.  Both the builds was successfull.  And these are
> some results of 'make check':

I would not doubt this is the -fomit-frame-pointer exception handling
bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
See PR 17220 and 6764.

-- Pinski


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  4:06 ` Andrew Pinski
@ 2004-12-22  4:34   ` Denis Zaitsev
  2004-12-22  4:38     ` Andrew Pinski
  2004-12-22 20:13   ` Denis Zaitsev
                     ` (2 subsequent siblings)
  3 siblings, 1 reply; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-22  4:34 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc, linux-gcc

On Tue, Dec 21, 2004 at 11:06:13PM -0500, Andrew Pinski wrote:
> 
> On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:
> 
> > I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
> > I mean either of these flags is present in both the exported shell
> > variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
> > from gcc/Makefile.  Both the builds was successfull.  And these are
> > some results of 'make check':
> 
> I would not doubt this is the -fomit-frame-pointer exception handling
> bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
> See PR 17220 and 6764.

Ok, I'll try.  Am I understand right, that the versions you've listed
are still unreleased?

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  4:34   ` Denis Zaitsev
@ 2004-12-22  4:38     ` Andrew Pinski
  0 siblings, 0 replies; 8+ messages in thread
From: Andrew Pinski @ 2004-12-22  4:38 UTC (permalink / raw)
  To: Denis Zaitsev; +Cc: gcc, linux-gcc


On Dec 21, 2004, at 11:34 PM, Denis Zaitsev wrote:

> On Tue, Dec 21, 2004 at 11:06:13PM -0500, Andrew Pinski wrote:
>>
>> On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:
>>
>>> I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
>>> I mean either of these flags is present in both the exported shell
>>> variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
>>> from gcc/Makefile.  Both the builds was successfull.  And these are
>>> some results of 'make check':
>>
>> I would not doubt this is the -fomit-frame-pointer exception handling
>> bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
>> See PR 17220 and 6764.
>
> Ok, I'll try.  Am I understand right, that the versions you've listed
> are still unreleased?

That is correct which is why I also pointed to the bug reports
which talk about the problem a little further but since the fix
is in all of the next releases I am not worried about the problem in
general.

-- Pinski


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  4:06 ` Andrew Pinski
  2004-12-22  4:34   ` Denis Zaitsev
@ 2004-12-22 20:13   ` Denis Zaitsev
  2004-12-22 22:20     ` Denis Zaitsev
  2004-12-27 21:14   ` Denis Zaitsev
  2004-12-28  4:08   ` Denis Zaitsev
  3 siblings, 1 reply; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-22 20:13 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc, linux-gcc

On Tue, Dec 21, 2004 at 11:06:13PM -0500, Andrew Pinski wrote:
> 
> On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:
> 
> > I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
> > I mean either of these flags is present in both the exported shell
> > variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
> > from gcc/Makefile.  Both the builds was successfull.  And these are
> > some results of 'make check':
> 
> I would not doubt this is the -fomit-frame-pointer exception handling
> bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
> See PR 17220 and 6764.

I'm trying the patches.  Make check is still going and this should
take the couple more hours.  But it's already obvious that nothing has
changed into the something better...  I've a lot of failures for
libstdc++.  The most of them look like:


FAIL: 22_locale/num_get/get/wchar_t/1.cc execution test
FAIL: 22_locale/num_get/get/wchar_t/2.cc execution test
FAIL: 22_locale/num_get/get/wchar_t/3.cc execution test
FAIL: 22_locale/num_get/get/wchar_t/5.cc execution test
FAIL: 22_locale/num_get/get/wchar_t/6.cc execution test
FAIL: 22_locale/num_get/get/wchar_t/9.cc execution test


BTW, as I can see, the test sources are compiling w/o
-fomit-frame-pointer.  And the only optimization is -O2.  And it's
strange by itself.

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22 20:13   ` Denis Zaitsev
@ 2004-12-22 22:20     ` Denis Zaitsev
  0 siblings, 0 replies; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-22 22:20 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc, linux-gcc

On Thu, Dec 23, 2004 at 01:13:27AM +0500, Denis Zaitsev wrote:
> I'm trying the patches.  Make check is still going and this should
> take the couple more hours.

Ok, it's finushed.  The libstdc++ results are:


# of expected passes            2366
# of unexpected failures        355
# of expected failures          7


There are 2 less failures...

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  4:06 ` Andrew Pinski
  2004-12-22  4:34   ` Denis Zaitsev
  2004-12-22 20:13   ` Denis Zaitsev
@ 2004-12-27 21:14   ` Denis Zaitsev
  2004-12-28  4:08   ` Denis Zaitsev
  3 siblings, 0 replies; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-27 21:14 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc, linux-gcc

On Tue, Dec 21, 2004 at 11:06:13PM -0500, Andrew Pinski wrote:
> 
> On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:
> 
> > I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
> > I mean either of these flags is present in both the exported shell
> > variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
> > from gcc/Makefile.  Both the builds was successfull.  And these are
> > some results of 'make check':
> 
> I would not doubt this is the -fomit-frame-pointer exception handling
> bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
> See PR 17220 and 6764.

And these are the results for GCC 3.4-20041224:


                === libstdc++ Summary ===

# of expected passes            2373
# of unexpected failures        357
# of expected failures          7


So, there are still the same bugs.

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++
  2004-12-22  4:06 ` Andrew Pinski
                     ` (2 preceding siblings ...)
  2004-12-27 21:14   ` Denis Zaitsev
@ 2004-12-28  4:08   ` Denis Zaitsev
  3 siblings, 0 replies; 8+ messages in thread
From: Denis Zaitsev @ 2004-12-28  4:08 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc, linux-gcc

On Tue, Dec 21, 2004 at 11:06:13PM -0500, Andrew Pinski wrote:
> 
> On Dec 21, 2004, at 10:42 PM, Denis Zaitsev wrote:
> 
> > I've built GCC with -fomit-frame-pointer and -fno-omit-frame-pointer.
> > I mean either of these flags is present in both the exported shell
> > variables CFLAGS and CXXFLAGS and also in the BOOT_CFLAGS variable
> > from gcc/Makefile.  Both the builds was successfull.  And these are
> > some results of 'make check':
> 
> I would not doubt this is the -fomit-frame-pointer exception handling
> bug which was fixed in 3.4.4, 3.3.6, and 4.0.0.
> See PR 17220 and 6764.

And these are the results for GCC 3.4-20041224:


                === libstdc++ Summary ===

# of expected passes            2373
# of unexpected failures        357
# of expected failures          7


Should I submit the new PR so?

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2004-12-28  4:08 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-12-22  3:42 [BUG] GCC 3.4.3/x86: -fomit-frame-pointer cause a lot of test failures for libstdc++ Denis Zaitsev
2004-12-22  4:06 ` Andrew Pinski
2004-12-22  4:34   ` Denis Zaitsev
2004-12-22  4:38     ` Andrew Pinski
2004-12-22 20:13   ` Denis Zaitsev
2004-12-22 22:20     ` Denis Zaitsev
2004-12-27 21:14   ` Denis Zaitsev
2004-12-28  4:08   ` Denis Zaitsev

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).